Understanding Clinical Trial Design - Research Advocacy Network

UNDERSTANDING CLINICAL TRIAL DESIGN: A TUTORIAL FOR RESEARCH ADVOCATES
Patient Allocation
Adaptive Design
Example
Patient Allocation Adaptive Design
In the patient allocation adaptive design an adaptation rule modifies the way
patients are assigned to intervention arms as a result of accumulating data.
Patients are always assigned randomly; what is modified is the proportion of
patients assigned to each intervention arm. In particular, more patients are
assigned to the intervention that is performing better. If the apparent advantage is
real, the intervention arms rapidly diverge and the trial can be concluded.
However, if the advantage is due to chance, the intervention arms converge. The
total number of patients needed to come to a reliable conclusion is determined as
data accumulate, but it is typically fewer than with a traditional one stage design.
Additionally, compared to the traditional randomized controlled trial, a larger proportion
of patients in the patient allocation adaptive design are treated with the
superior intervention. Don Berry, Ph.D. of MD Anderson Cancer Center has
used this design effectively in a number of applications, and his simulations
demonstrate its efficiency in finding “winners” among a group of potential interventions.
Figure 14 (page 30) presents an example of a two-arm trial in which assignment
of patients to intervention is modified through the course of the trial.
The adaptation rule specifies that at the start of the trial patients should be
allocated equally to the two treatment arms. The adaptation rule also indicates
when data should be analyzed, and depending on the results, how the
allocation ratio should be changed.
In this example, data were analyzed after outcomes from the first eight
patients became available. The response rate (i.e., percentage of patients
whose tumors shrink following treatment) was twice as large in one of the
arms. The adaptation rule specified that in such cases, the ratio of patient
assignment should change from 1:1 to 3:1, favoring the superior arm.
As more data are collected outcomes would continued to be compared at
pre-specified points. The adaptation rule would specify under which condition
the trial should be:
1) terminated due to clear superiority of one treatment;
2) ended due to clear equivalence of the treatments (i.e., futility of finding
difference)
3) continued, with the same patient allocation; or
4) continued with a revised patient allocation.
Note that throughout the trial, all patients were randomly assigned to the
experimental and control arms based on a randomization algorithm, even
though the proportion of patients assigned to each arm shifted through the
course of the trial.
29